N1 (9) January-March 2006 CONTENTS

Bucha V.V., Ablameyko S.V. Algorithm of interactive image segmentation based on dynamic programming (p.5)

Interactive segmentation technique based on dynamic programming and fast search in hash table is proposed. The proposed algorithm differs from published work in speed and robustness. It allows extracting contour and skeleton as well. The proposed algorithm was tested on remote sensing images for segmentation task. In comparison with hand digitization and semiautomatic approaches the results of segmentation of line and area cartographic objects are achieved in more convenient and accurate way.

Zhuk D.V., Tuzikov A.V. Three-dimensional model reconstruction using two digital images (p.16)

The paper surveys a process of a three-dimensional model reconstruction using a pair of digital images. The subtasks of the problem including automatic finding of corresponding points, image rectification, dense disparity map computation and triangulation are discussed. Algorithms used at each step of the reconstruction process and some their modifications are described.

Zalesky B.A., Lukashevich P.V. Reconstruction of 3D surfaces by spline technique (p.27)

A one-parameter spline technique is presented in order to reconstruct 3D surfaces in raster or vector forms from several closed, in general, non-planar curves, including contours that were outlined manually. The techniques allow visualization of 3D objects and evaluating their volumes without reconstruction of 3D image cube. A method to smooth rough or noisy 3D surfaces is also described. The local parametrization of surfaces is exploited to reduce the problem to the case of smoothing of functions of two variables.

Kiselev V.V. Computer model of allophone-wave synthesis of russian speech (p.36)

The system of speech synthesis is described based on allophone-wave presentation of a signal. General scheme of the system is given. Functional features of speech synthesizer module components are described. Special attention is devoted to preprocessing stage of text processing. Linguistic, phonetic, prosodic and acoustic blocks of Russian language synthesizer are described. The peculiarity of the system is the modularity and multilayer structure to realize multilingual and multivoiced speech synthesis by text.

Davydau A.G. Dynamic programming speech segmentation algorithm (p.47)

A system of automatic speech segmentation is considered on the basis of dynamic programming. The spectrum and averaged spectrum final differences on time are offered as a feature vector. The optimum parameters of the system are determined on the test set including 1128 elements.

Bibilo P.N. Comparison of circuit implementations of VHDL models of incompletely specified Boolean functions (p.58)

In the paper VHDL models of incompletely specified Boolean functions and systems of such functions are suggested. Using the models of incompletely specified functions allows to obtain less complicated logical circuits in LeonardoSpectrum synthesizer as compared with the models resulting in completely specified functions. Complementary using Custom Logic synthesis system allows to organize one more design route by transition to two-level representations of implemented functions.

Zakrevskij A.D., Toropov N.R. Solving the majority of linear logical equations of inconsistent system on supercomputer (p.68)

Overdefined systems of linear logical equations are considered, the number of equations in which surpasses the number of variables. As a rule, such a system has no root and therefore is called inconsistent. Anyhow, these systems can be solved in some sense. For example, from the cryptography point of view, it is interesting to find solutions satisfying the maximum number of equations or, if the right parts are distorted, to restore the system. The parallel implementation of a randomized algorithm is suggested for solving inconsistent systems on the supercomputer SKIF. The results of experiments testify the efficiency of parallel computations when solving large systems of linear logical equations.

Kochanov D.A. VHDL models of mathematical functions that are based on piecewise-polynomial interpolation (p.78)

This article proposes a solution of a problem of creating VHDL-models of macroelements, which are designed to calculate the values of analytical functions. The initial function definition is given in analytical form in mathematical packages. The resulting function description is provided in table form with conversion to VHDL form.

Leonchik P.V. Minimization of Boolean function systems in the class of disjunctive normal forms (p.88)

The paper presents the DNF-minimization programme of systems of fully specified Boolean functions. The comparative analysis of the effectiveness of the developed algorithm realized in the programme Tie and the programme Espresso is presented. The results of statistical experimental testing of the algorithms on pseudorandom Boolean functions and on a wide set of Benchmark examples are given.

Murashko I.A. Power consumption analysis of XOR based circuits (p.97)

This paper is dealing with power consumption of XOR based circuits. T he upper and lower bounds for switching activities of modulo two circuits have been obtained. The algorithm for XOR based circuit synthesis with minimal switching activity is offered.

Yarmolik S.V., Yarmolik V.N. Memory pattern sensitive faults detection using mulpiple runs of march tests (p.104)

This paper deals with memory pattern sensitive faults detection problem. It shows the efficiency of using march tests for detection memory passive sensitive faults. It shows also the efficiency of multiple runs of march tests for detection PPSF and analyzes the optimal address seeds. At the conclusion the experimental data, showing the fault coverage of multiple uses of march tests are given.

Kartynnick A.V., Linkevich A.D. A global attractor and sufficient conditions for retrieval of patterns by neural networks of Cohen - Grossberg (p.114)

The Cohen - Grossberg model of analogous neural networks is studied in the general case when interneuronal (synaptic) connections can be non-symmetric and consequently there is no usual analogue of the Hamiltonians for the systems in complete phase space. It is proved that under certain conditions there exists a global attracting region Q so that all attractors of the system lie inside the Q and, besides, the domain Q is an attractor of the system as well. The upper bound has been estimated for time to reach a prescribed neighbourhood of the region Q . Sufficient conditions have been found under which the network converges asymptotically to a nearest fixed point (memorized pattern) and reaches a given neighbourhood of this attractor in a certain time. An upper bound is given for time needed to retrieve a memorized pattern with a prescribed accuracy.

Tarasevich A.A. Minimizing total penalty for sheduling jobs with non-zero release dates under a common due date based on tabu search (p.124)

In the paper one can find a tabu search heuristic algorithm for scheduling jobs with non-zero release dates on a single machine to minimize the total penalty for early and tardy completion of jobs and assignment of a common due date.

Prokhorova A.A. Abilities and tools of reengineering in contemporary designing systems (p.130)

The task of space model designing on the basis of imported model without history is considered . Decisions of the task in some space modeling systems are shown.

The program of Belarus-Russian Union State: «The development and implementation in the states-participants of the Union State of science intensive computer technologies on the basis of multiprocessor computing systems» (code «TRIADA») (p.136)

CONFERENCES

The Fifth international conference «Information processing and control in extreme situations» (IPCES'2006) (p.139)
Site of conference >>>